Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A system for performing program analysis, comprising: a first logical extractor software feature configured to receive a plurality of programs of a first platform that have been assigned a first label, receive a plurality of programs of the first platform that have been assigned a second label, convert each of the plurality of programs of the first platform that have been assigned the first label to a first set of platform-independent logical features, in a platform-independent form, that are universal to both the first platform and a second platform, wherein the first platform and the second platform represent different operating systems, and convert each of the plurality of programs of the first platform that have been assigned the second label to a second set of platform-independent logical features in the platform-independent form, that are universal to both the first and second platforms; a computer learning trainer software feature configured to train a discriminatory model or classifier, using machine learning, based on the conversion of the programs of the first platform that have been assigned the first and second label as the first and second sets of platform-independent logical features, to distinguish between programs of the first label and programs of the second label; a second logical extractor software feature configured to receive an unlabeled program of the second platform, and convert the unlabeled program to a third set of platform-independent logical features, of the platform-independent form; a software tool, executed on a computer system, for analyzing a program, configured to use the trained discriminatory model or classifier to determine if the unlabeled program warrants the first label or the second label, based on the conversion of the unlabeled program as the third set of platform-independent logical features, after the unlabeled program has been converted to the third set of platform-independent logical features in the platform-independent form; and a mobile electronic device, of the second platform, having one or more applications installed thereon, wherein the mobile electronic device is configured to have the software tool analyze each of the plurality of applications stored thereon, wherein each of the platform-independent logical features represents a single program feature including an API call, a library used, a manner in which the Internet is accessed, or another program capability that has been converted from a platform-dependent program element.
This system performs cross-platform program analysis using machine learning. It includes a first component that takes programs from a first operating system (OS) with assigned labels (e.g., "malicious," "benign") and converts them into "platform-independent logical features." These features are universal, applying across different OSs, like API calls, used libraries, or how the program accesses the internet. A second component then uses these converted features to train a machine learning model that distinguishes between the different labels. A third component takes an unlabeled program from a *second* OS (e.g., a mobile device application) and converts it into the same platform-independent features. Finally, a software tool uses the trained machine learning model to determine the label for this unlabeled program, effectively classifying it. This entire process happens on a computer system, often analyzing applications on a mobile electronic device, which represents the second platform.
2. The system of claim 1 , thither comprising an application repository for storing, a plurality of applications for the second platform, wherein the application repository is configured to have the software tool analyze each of the plurality of applications stored therein.
This system, which performs cross-platform program analysis using machine learning, takes labeled programs from a first operating system (OS), converts them into universal platform-independent logical features (like API calls or used libraries), and uses these features to train a machine learning model that distinguishes between labels. It then takes an unlabeled program from a *second* OS, converts it into the same features, and uses the trained model to classify it. This extended system further includes an application repository. This repository stores multiple applications specifically for the *second* operating system, and the system's software tool analyzes each of these applications directly from the repository.
3. The system of claim 1 , wherein at least some platform-independent logical features of the first set of platform-independent logical features, the second set of platform-independent logical features, and the third set of platform-independent logical features are indications of what hardware resources are called.
This system, which performs cross-platform program analysis using machine learning, takes labeled programs from a first operating system (OS), converts them into universal platform-independent logical features (like API calls, used libraries, or internet access patterns), and uses these features to train a machine learning model that distinguishes between labels. It then takes an unlabeled program from a *second* OS, converts it into the same features, and uses the trained model to classify it. A specific aspect of this system is that at least some of these platform-independent logical features, whether derived from the initial training programs or the unlabeled programs being analyzed, specifically represent indications of which hardware resources the programs access or call.
4. A computer system comprising: a processor; and a non-transitory, tangible, program storage medium, readable by the computer system, embodying a program of instructions executable by the processor to perform method steps for performing program analysis, the method comprising: receiving a plurality of programs of a first platform that have been assigned a first label; receiving a plurality of programs of the first platform that have been assigned a second label; converting each of the plurality of programs of the first platform that have been assigned the first label to a first set of platform-independent logical features, in a platform-independent form, that are universal to both the first platform and a second platform, wherein the first platform and the second platform represent different operating systems; converting each of the plurality of programs of the first platform that have been assigned the second label to a second set of platform-independent logical features, in the platform-independent form; training a discriminatory model or classifier, using machine learning, based on the conversion of the programs of the first platform that have been assigned the first and second label as the first and second sets of platform-independent logical features, to distinguish between programs of the first label and programs of the second label; receiving an unlabeled program of the second platform; converting the unlabeled program to a third set of platform-independent logical features, in the platform-independent form; and using the trained discriminatory model or classifier to determine if the unlabeled program warrants the first label or the second label, based on the conversion of the unlabeled program as the third set of platform-independent logical features, after the unlabeled program has been converted to the third set of platform-independent logical features in the platform-independent form, wherein each of the platform-independent logical features represents a single program feature including an API call, a library used, a manner in which the Internet is accessed, or another program capability that has been converted from a platform-dependent program element.
This describes a computer system executing a method for cross-platform program analysis using machine learning. The system first receives multiple programs from a *first operating system (OS)*, some pre-assigned a "first label" and others a "second label." It then converts each of these programs into "platform-independent logical features." These features are in a universal format applicable to both the first OS and a *second OS* (different platforms), and include elements like API calls, libraries used, or how the program accesses the internet. Next, the system trains a machine learning model using these converted features to learn how to distinguish between programs with the first and second labels. After training, the system receives an unlabeled program from the *second OS*, converts it into the same platform-independent features, and then uses the trained machine learning model to determine if the unlabeled program should be assigned the first or second label, effectively classifying it.
5. The computer system of claim 4 , further comprising: receiving a plurality of programs of the second platform that have been assigned the first label; receiving a plurality of programs of the second platform that have been assigned the second label; converting each of the plurality of programs of the second platform that have been assigned the first label to a fourth set of platform-independent logical features; and converting each of the plurality of programs of the second platform that have been assigned the second label to a fifth set of platform-independent logical features, wherein the training of the discriminatory model or classifier further includes using machine learning; based on the conversion of the programs of the second platform that have been assigned the first and second label as the fourth and fifth sets of platform-independent logical features, to distinguish between programs of the first label and programs of the second label.
This computer system executes a method for cross-platform program analysis using machine learning. It initially receives labeled programs from a first operating system (OS), converts them into universal platform-independent features (like API calls or libraries), and uses these to train a machine learning model to distinguish labels. It then receives and classifies unlabeled programs from a second OS using this model. This enhanced method further includes receiving additional labeled programs, but this time from the *second OS* itself, with programs assigned the "first label" and others the "second label." These programs from the second OS are also converted into their own sets of platform-independent logical features. The machine learning model's training process is then expanded to incorporate *all* these converted features – from both the first and second operating systems – to improve its ability to distinguish between programs assigned the first and second labels.
6. The computer system of claim 4 , wherein the first label represents malicious programming and the second label represents benign programming.
This computer system executes a method for cross-platform program analysis using machine learning. It receives labeled programs from a first operating system (OS), converts them into universal platform-independent features (like API calls or libraries), and uses these to train a machine learning model. It then receives and classifies unlabeled programs from a second OS using this model. A specific aspect of this method is that the "first label" used for classification represents *malicious programming*, while the "second label" represents *benign programming*, allowing the system to identify potential threats or safe software across different operating systems.
7. The computer system of claim 5 , additionally comprising: receiving a plurality of programs of a third platform, different from the first and second platforms, that have been assigned the first label; receiving a plurality of programs of the third platform that have been assigned the second label; converting each of the plurality of programs of the third platform that have been assigned the first label to a sixth set of platform-independent logical features; and converting each of the set of programs of the third platform that have been assigned the second label to a seventh set of platform-independent logical features, wherein the training of the discriminatory model or classifier further includes using machine learning, based on the conversion of the programs of the third platform that have been assigned the first and second label as the sixth and seventh sets of platform-independent logical features, to distinguish between programs of the first label and programs of the second label.
This computer system performs cross-platform program analysis by training a machine learning model and then classifying unlabeled programs. The method involves first receiving labeled programs from a first operating system (OS) and a second OS, converting all these programs into universal platform-independent features (such as API calls or used libraries). The system then trains a machine learning model using these features to distinguish between programs based on their assigned labels. After training, it uses this model to classify unlabeled programs from the second OS. This enhanced method further expands its training data by additionally receiving labeled programs from a *third operating system*, distinct from both the first and second. These third-platform programs, also categorized with first and second labels, are similarly converted into platform-independent features. The machine learning model's training process is then extended to incorporate all these converted features from the first, second, *and third* operating systems, significantly broadening the model's ability to accurately distinguish between programs assigned the first and second labels across an even wider range of platforms.
8. The computer system of claim 7 , further including, receiving a plurality of programs of the first platform that have been assigned a third label; and converting each of the plurality of programs of the first platform that have been assigned the third label to an eighth set of platform-independent logical features, wherein the training of the discriminatory model or classifier further includes, using machine learning, based on the conversion of the programs of the first platform that have been assigned the first, second, and third labels, to the first, second, and eighth sets of platform independent logical features, to distinguish between programs of the first, second, and third labels.
This computer system performs cross-platform program analysis by training a machine learning model on programs from multiple platforms and then classifying unlabeled programs. The method involves receiving labeled programs from first, second, and third operating systems (OS), converting all these programs into universal platform-independent features (like API calls or used libraries), and then training a machine learning model to distinguish between programs based on their assigned "first" and "second" labels. It then classifies unlabeled programs from the second OS. This further refined method additionally receives programs from the *first OS* that have been assigned a *third label*. These programs are also converted into platform-independent features. The machine learning model's training process is then further expanded. Instead of just distinguishing between two labels, the model is now trained to distinguish between programs assigned the *first, second, and third labels*. This training uses all the converted platform-independent features from the first, second, and third OSs, for all three labels, enhancing the model's ability to categorize programs into three distinct classes based on their cross-platform features.
9. The computer system of claim 4 , wherein at least some platform-independent logical features of the first set of platform-independent logical features, the second set of platform-independent logical features, and the third set of platform-independent logical features are indications of what hardware resources are called.
This computer system executes a method for cross-platform program analysis using machine learning. It receives labeled programs from a first operating system (OS), converts them into universal platform-independent features (like API calls or libraries), and uses these to train a machine learning model. It then receives and classifies unlabeled programs from a second OS using this model. A specific aspect of this method is that at least some of these platform-independent logical features, whether extracted from the initial training programs from the first OS or from the unlabeled programs being analyzed from the second OS, specifically represent indications of what hardware resources the respective programs access or call during their execution.
Unknown
July 28, 2020
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.